Recognition of Actual Grip Force Movement Modes Based on Movement-related Cortical Potentials

Abstract

A new paradigm of grip force movement with parameters involving right and left hands is put forward in the study to meet the needs of brain-computer interface based brain-machine interaction control(BMIC) — direct brain-controlled robot interface(BCRI). Time-domain feature representation for grip force movement-related cortical potentials/movement-related potentials(MRPs) and the single-trial recognition of grip force movement modes are explored under the paradigm. EEG signals were picked up from eleven healthy subjects during four different tasks of right and left hands. Subjects were asked to execute voluntary grip movement at two modes of grip force variation. Each task was executed 30 times in a random order repeatedly. The features having significant difference among different grip force tasks are used for the classification of grip force modes by Fisher linear discrimination analysis based on kernel function(k-FLDA) and support vector machine(SVM), respectively. The study further demonstrates that MRPs may reflect brain neural mechanism process for planning, execution and precision of a given grip movement task. The average misclassification rates of 24 ± 4 % and 21 ± 5 % across eleven subjects are achieved by k-FLDA and SVM, respectively.The minimum misclassification rate is 12 % and the average of minimum misclassification rates across eleven subjects is20.9 ± 5 %. The study is expected to lay a foundation for follow-up comparative researches, which provide some additional force control intention instructions for BMIC/BCRI.